From edge to extreme edge: Exploring the dimensions of AI inference in IoT

This talk presents an in-depth exploration of the dual facets of AI inference within the IoT ecosystem: the established edge computing and the emerging "extreme edge", where TinyML is gaining prominence. We delve into how these two spheres interact and complement each other, driving advancements in AI-driven IoT applications. The discussion highlights how the proliferation of AI inference accelerators and varying software, hardware, and networking configurations are shaping the landscape of computational efficiency and real-time responsiveness. We will explore the development of frameworks that accommodate hardware diversity, ensuring optimized AI inference performance across a spectrum of devices. The seminar will also shed light on the strategies for seamless integration and coordination of devices at both the edge and the extreme edge. Emphasizing the importance of data-driven decision-making, we will discuss how these technologies enable robust AI capabilities in diverse applications. As a future perspective, we will briefly consider the potential role of emerging AI technologies, such as generative AI, in further enhancing the capabilities and applications of AI inference within IoT.

 Link: https://www.youtube.com/watch?v=2_G1Emaj4qg